In recent years there has been much excitement about genetically encoded fluorescent indicators of neural activity, with new molecules such as the genetically encoded calcium indicator GCaMP6 being used to image the activity of many neurons at once in living brains. However, such indicators are slow, raising the question of whether voltage indicators will become useful enough to be widespread in neuroscience. Furthermore, imaging of axons and dendrites remains difficult, especially in densely expressing tissues. For example, when neurons express such reporters densely, axons and dendrites within the diffraction limit of light will have their signals mixed, so that the signals of individual neural processes cannot be resolved. How can we push the spatiotemporal performance of neural activity imaging to the specifications desired by neuroscientists ? down to the millisecond timescale, and down to the sub-micron scale axonal and dendritic parts of neurons? We here propose to address this problem through molecular engineering, guided by in vivo imaging constraints. To address the spatial dimension: if neural activity indicators could be safely clustered into discrete, bright puncta that, even when expressed in all the cells of a neural circuit, are separated from one another by a distance greater than the diffraction limit of the imaging system, then these puncta could cleanly be imaged, and used to sample activity along axons and dendrites of the neurons in a circuit. In this grant, we will (Aim 1) create and validate this strategy, which we call stochastic arrangement of reagents in clusters (STARC). In this way, we will effectively point the way towards circuit-wide neural activity imaging that allows for the investigation of axonal signaling and dendritic processing, and not only cell body imaging. To address the temporal dimension: we will create optimized fluorescent voltage indicators (Aim 2). Pioneering efforts have resulted in fluorescent voltage indicators, but their performance is often poor when utilized in the brain, because of poor trafficking and membrane localization that manifests in vivo, since neurons in vivo are different from the cultured cells used to screen for the voltage sensors. We will conduct an in situ screen to directly identify fluorescent voltage indicators that work well in neurons in intact mouse brain circuits, by virally expressing members of a library of mutant voltage indicators directly in the mouse brain, imaging the responses with single cell resolution in mouse brain slices, and then directly reading out the mutations that yielded the voltage indicators that best perform in actual brain circuits, validating the resultant indicators in the mouse brain. We will also create (Aim 3) STARC forms of voltage sensors, since the proximity issues discussed in Aim 1 are even more severe when a neural activity reporter is on a neural membrane that is in close proximity to other membranes. We will close the loop by testing all such indicators in vivo and then iterating on the molecular engineering, delivering to the neuroscience community a powerful, simple-to-use toolbox that can be rapidly deployed for ultraprecise ? across both space (via STARC) and time (via in situ optimized voltage indicators) -- neural activity imaging.